No omniscience. The clues are laid out, possibilities enumerated, implausible options pushed down, and the survivors given a probability, a basis and a test.
Enumerate before you eliminate. People typically latch onto the first explanation that comes to mind. Ockham forces the alternatives onto the table first, then filters — which counters anchoring and confirmation bias.
Let base rates temper the story. A vivid explanation can still be very unlikely if such things rarely happen. That is Bayesian thinking: look at how common it is, then at how strong the evidence is.
Give a way to verify, not a verdict. Every high-probability option comes with a concrete test — what to observe, record or compare — until the clue supports or refutes it.
Occam's razor. When two explanations fit the available clues, keep the one with fewer assumptions and better testability. New clues may overturn the previous conclusion — that is the point.